Organización de las actividades académicas del programa

Across all cases, the CDK2/7/9 inhibitor SNS-032 was the most effective substance. To further elucidate the effect of SNS-032 on the most central growth-regulating receptor in T-cells, namely the TCR, Jurkat T-cells containing an NFAT-coupled luciferase reporter to read out the activation of this distal TCR signaling effector were treated with SNS-032 in the presence of TCR crosslinking anti-CD3 and anti-CD28 antibodies. There was an NFAT-inhibition down to 45% in the presence of 1µM SNS- 032 (Fig 24A). According to our gene expression data, the TCR signaling pathway in T-PLL cells at baseline appears to be activated by the overexpression of PI3K,

MEKK1, Ras, and NFAT in addition to the downregulation of inhibitory signals by CTLA4 (Fig 24B).

A.

B.

Figure 24. A. Jurkat NFAT luciferase reporter cells were incubated with the indicated concentrations

of SNS-032 on a 384-well plate for 5 hours in the presence of purified anti-CD3 and anti-CD28 antibodies. B. TCR signaling pathway from IPA software overlaid with gene expression data from RNA sequencing of 4 PLL patient samples and 2 CD4+ healthy control samples.

0 0.5 0 100 200 300 400 500 anti-CD3/anti-CD28 (μg/ml) Luminescence 0.0010 0.01 0.1 1 10 25 50 75 100 SNS-032 (µM) 5 h NF A T activity %

Discussion

The aim of studies I–III was to characterize STAT-mutation negative LGL leukemia molecularly through the use of next-generation sequencing and gene expression analysis. Our previous findings revealed somatic mutations in the SH2-domain of either the STAT3 or STAT5B genes in approximately 40–50% of LGL leukemia cases2,19. However, as the STAT-mutation negative patients also harbor large

expansions of LGL cells, other somatic mutations may drive the development of the disease.

In study I, we showed that mutations in genes connected to the STAT-pathway and T- cell activation/proliferation represent rare genetic triggers for T-LGL leukemia. Previous studies have shown that almost all T-LGLL patients exhibit the activation of STAT3, proving that this pathway is essential for the pathogenesis of LGL

leukemia34,2,35. In this study, we also found that STAT3 was phosphorylated in patients without actual STAT3 mutations, indicating that other underlying factors are

leading to the activation of the pathway. To further confirm this finding, T-LGLL

patients were also clustered independently of their mutational status in the expression analysis. Therefore, the novel V995M mutation in the PTPRT gene is particularly interesting, as it may directly impact STAT3 activation and is the first evidence of other mutations affecting the JAK/STAT pathway in LGL leukemia.

PTPRT is known to reverse the phosphorylation of Tyr705 on STAT3, a modification that is associated with STAT3 deactivation70_{. The PTPRT V995M mutation may}

thereby affect STAT3 activity by reducing the dephosphorylation of Tyr705, leading to an increased expression of STAT3 target genes. The mutation was found within the phosphatase domain D1 that is responsible for the proteins’ phosphatase activity.

PTPRT is also frequently disabled in other cancers, such as lung and gastric cancer,

where the hypermethylation of the promoter of PTPRT increases STAT3 activation and causes sensitivity to STAT3 inhibition165,166. The effect of other missense mutations in the D1 domain of PTPRT have been studied167 and the mutant proteins show a decrease in thermal stability and activation energy for phosphatase activity with respect to the WT protein. Inactivating mutations of the PTPRT gene may thereby have the same functional consequence as activating mutations of STAT3 in T- LGLL patients.

To clonally expand, T-cells must have an acquired survival advantage. In addition to defects in apoptotic pathways, survival pathways such as PI3K-AKT may also be activated in LGL leukemia patients34. Therefore, it was interesting to observe novel missense mutations in BCL11B and NRP1 in T-LGLL patients. BCL11B functions as a transcription factor required for normal T-cell development. The inactivation of

BCL11B in mice leads to thymocyte developmental arrest and aberrant self-renewal

cells results in apoptosis171. Mutations in BCL11B have also been reported in 9–16% of T-ALL patients22,172,173.

NRP1, a receptor involved in axon guidance, is also expressed in human DCs and resting T-cells158. Interestingly, earlier studies have shown that NRP1 can play a role in the primary immune response during the formation of the immunological synapse between DCs and resting T-cells. The preincubation of DCs and resting T-cells with blocking NRP1 antibodies inhibited the DC-induced proliferation of T-cells. Thus the mutation in the NRP1 gene may affect the proliferation capacity of T-cells.

These novel mutations affecting either the STAT3 or T-cell activation pathway were not found to be recurrent in our screening cohort (n=113). It seems that the STAT- mutation negative patients are a more heterogeneous patient cohort, and therefore no similar dominant mutation as in STAT3 will be found in this group. However, in the Sanger sequencing screening assays, the primers covered only the mutation spots and nearby base pairs, and it is possible that by screening the whole genes, some

additional mutations in the same genes will be found.Additionally, low frequency

variants might go undetected due to the poor sensitivity of Sanger sequencing.To date, our group has performed the exome sequencing of 20 LGLL patients, and few recurrent mutations have been discovered. As mutations occurring in different genes with similar functions can drive a similar phenotype, further bioinformatical

investigation of the different somatic variants should be performed to highlight pathways activated by mutations other than STAT3/5B in LGL leukemia.

In studies II and III, additional STAT-mutations were revealed in the process of the exome sequencing of T-LGLL patients. First, a subgroup of CD8+ T-LGLL patients (4%) harboring mutations in the DNA-binding and coiled-coil domain of STAT3 were discovered. STAT3 activation is traditionally thought to be mediated through the dimerization interface containing the SH2-domain174. While the constitutive

activation of STAT3 in LGL leukemia patients with SH2 hotspot mutations could be explained by an increased stabilization of STAT3 homodimers through enhanced hydrophobic attraction, the mode of activation through other domains is not as straightforward. Arginine residues within the DNA binding domain of STAT3 have, however, been found to promote phosphorylation and the intracellular shuttling of STAT3175

. An arginine–glutamine exchange at the STAT3 moieties R414 and R417 reduces the cytokine-dependent tyrosine phosphorylation of STAT3. The H410R variant seen in two LGL leukemia patients results in one extra arginine residue in this area, and it is plausible that this increase in hydrophilicity within the DNA binding domain mediates the activation of STAT3. The H410R mutation was also seen to be recurrent—one of the hallmarks of activating mutations.

Previously, germline mutations in the DNA binding domain of STAT3 have been found in different immune disorders. In hyper IgE syndrome 176, dominant-negative mutations in the DNA binding domain leads to the diminished DNA binding ability of

the STAT3 dimer. There has also been some interest in using specific STAT3 inhibitors in the treatment of other hematological malignancies, and recently a small molecule compound specifically targeting the STAT3 DNA-binding domain was found to inhibit cancer cell proliferation, migration, and invasion177.

One CD8+ T-LGLL patient was found to harbor an F174S mutation in the coiled-coil domain of STAT3. This domain consists of four antiparallel helices and has previously been shown to be essential in receptor binding mediated by the SH2-domain as well as subsequent activation. The α1 region, containing the F174S mutation, seems to be crucial for STAT3 tyrosine phosphorylation stimulated by the epidermal growth factor (EGF) and IL6 as the deletion of the α1 abolished tyrosine phosphorylation of STAT3 in EGF-induced COS-1 cells178

. One possible mechanism for the involvement of the coiled-coil domain in STAT3 phosphorylation is through the reinforcement of the protein-protein interaction between STAT3 and the activated IL6 receptor. Mutations in the coiled-coil domain have previously been seen in patients with inflammatory hepatocellular adenomas, where the E166Q mutant was shown to be constitutively phosphorylated on Tyr705, hypersensitive to IL6, and able to

translocate to the nucleus 179. Persistent STAT3 activation has also been found in activated B-cell-like diffuse large B-cell tumors, where the M206K STAT3 mutation leads to increased cell proliferation180.

In study III, exome sequencing identified novel somatic missense mutations in the transactivation domain of STAT5B in two CD4+ T-LGLL patients. Altogether, the

STAT5B mutation frequency in the CD4+ T-LGL leukemia patients in our cohort was

55% (6/11 patients). This is significantly higher than in the previous study of 211 CD8+ T- and NK-cell LGL-leukemia cases, where STAT5B SH2 domain mutations were initially discovered in 2% of the cases19_{. The reported STAT5B-mutated patients}

had a much more aggressive disease course than that of typical T-LGLL. This finding, together with the high prevalence of STAT5B mutations in aggressive T-cell

neoplasms, such as T lymphoblastic leukemia103,104, T-PLL4, and hepatosplenic T-cell lymphoma95, may suggest that STAT5B mutations predict a more aggressive clinical behavior than does the mutational activation of STAT3. However, in our cohort of

STAT5B-mutated CD4+ T-LGLL patients, the disease course was quite indolent, and

none of the patients with STAT5B mutations needed treatment during the observation time. The analysis of STAT5 target genes with ChIPseq has shown that STAT5B binds to molecules such as DOCK8, SNX9, FOXP3, and IL2RA, making it a key factor in T-cell development181

. Together, these results suggest that STAT5B plays a central role in the development of T-cell neoplasms and that STAT5B mutations can be considered a novel diagnostic marker for CD4+ T-LGLL.

The diagnosis of T-LGL leukemia is often difficult, as symptoms and immuno- phenotypical findings can resemble other reactive conditions, making the discrimination between malignant lymphoproliferation and reactive processes challenging. While T-cell clonality is considered an important finding supporting diagnosis, its significance remains controversial, as a proportion of healthy elderly individuals also present with non-neoplastic clonal T-cell expansions182

. Therefore, consistent detection of STAT3 and STAT5B mutations in T-LGL leukemia represents a more definitive new diagnostic biomarker. Subsequently, the importance of screening the entire STAT3 and STAT5B genes in the diagnostic workup of T-LGLL is further emphasized.

There also seems to be a clear division between the occurrence of STAT3 mutations in T-LGLL CD8+ cases versus the occurrence of STAT5B mutations in CD4+ cases. What could be the biological reason for this division? For STAT3, it is known that the genes of the activated IL6-STAT3 axis are overexpressed in CD8+CD57+ LGL cells due to chronic immune activation in T-LGLL37

. Because the transcription of a gene has been found to correlate with the rate of mutagenesis183

, this could explain the increased number of somatic STAT3 mutations. As for STAT5B, the upstream regulator IL-15 has been found to stimulate the proliferation and survival of CD4+ memory T-cells184

. In a network model for T-LGLL survival signaling, IL-15 and PDGF were able to sustain a leukemic population185_{, and the overexpression of IL-15}

has been observed in T-LGLL, in autoimmune disorders, and during viral infections. Cytomegalovirus-derived stimulation has been associated with CD4+ T-LGL cases163_,

which may explain the higher STAT5B variant rate seen in CD4+ cases.

Another leukemic disease where CD4+ cells are frequently STAT5B-mutated is T- PLL. Although the mutation sites are overlapping, the resulting disease phenotypes are completely different, suggesting that in the more aggressive T-PLL, other

aberrations such as chromosomal translocations are also driving the disease. In study IV, we systematically explored the diversity of drug responses in a large collection of T-PLL patient samples ex vivo using a drug screening platform and explored the associations of drug sensitivities with genetic aberrations. Such a strategy has proven successful before in BCR-ABL-driven leukemias, where axitinib was found to be effective in patients harboring the resistance-causing T315I mutation in ABL1135. The deep amplicon sequencing of JAK/STAT pathway genes in our patient cohort revealed recurrent mutations in 69% of cases. Single-gene mutations affected JAK3 (29%), STAT5B (7%), JAK1 (6%), and IL2RG (1%), while coexisting multiple

mutations were found in 26% of the cases, either within more than one gene or within the same gene. Compared to the previous genetic profiling of JAK/STAT pathway genes in T-PLL patients4,51,52,186, these results were highly concordant. For example, Kiel et al. (2014) reported a 76% JAK/STAT mutation frequency in T-PLL (38/50). Overall, the recurrent involvement of JAK/STAT signaling elements in non-

synonymous mutations and indications of an activated JAK/STAT axis implicate a new actionable lesion to be explored in T-PLL. However, our sensitivity screen revealed that the ex vivo sensitivity of patient samples to JAK-inhibitors, such as ruxolitinib, momelotinib, tofacitinib and gandotinib, was not directly driven by these mutations in the JAK/STAT pathway. However, it was interesting to note that, for ruxolitinib, a JAK1/2 inhibitor, patients with the highly activating N642H STAT5B mutation were not as sensitive to the drug as patients with other STAT5B mutations. This could be due to the stability of the N642H mutant homodimer compared to the wild-type STAT5B (25-fold higher binding affinity), whereby the over-activation cannot be suppressed completely with JAK-inhibitors alone187.

Generally, the clustering of drug sensitivity values in T-PLL revealed three patient subgroups driven by their resistance/sensitivity to HDAC-, HSP90-, and

PI3K/mTOR/Akt-inhibitors. Surprisingly, despite the prevalence of the signature event of the activation of TCL1A (an established AKT coactivator) in the majority of T-PLL cases, only a subset responded to PI3K/AKT/mTOR inhibitors, such as AZD- 8055, MK-2206, apitolisib, and dactolisib. This might be due to the rather initiating role of TCL1A or its predominant effect in the context of TCR activation. On the other hand, the HDAC-inhibitors vorinostat, panobinostat, CUDC-101, quisinostat, and belinostat showed efficacy in the majority of PLL samples ex vivo. Recently, it was shown that treating relapsed or refractory PLL-patients with a combination therapy consisting of cladribine and an HDAC inhibitor re-sensitizes patients to alemtuzumab and results in better overall survival188. Further highlighting the importance of epigenetic regulation in PLL, mutations in genes encoding epigenetic regulators EZH2, TET2, and BCOR have been found in a number of T-PLL

patients189.

There was an outstandingly uniform sensitivity of virtually all cases to the CDK2/7/9- inhibitor SNS-032, which had been tested in phase I for advanced chronic

lymphocytic leukemia (CLL) and multiple myeloma (MM)190,191. These data are intriguing, since they indicate cell cycle dysregulation as a thus-far underappreciated central feature of T-PLL cells and implicate a novel vulnerability. As a potentially underlying genomic lesion, a deletion at 12p13 is found in approximately half of T- PLL patients and has been associated with haplo-insufficiency of the CDKN1B gene encoding the cyclin-dependent kinase inhibitory protein p27190. In our expression analysis, we also found that CDKN1A, a protein that usually inhibits the activity of CDK2 and CDK4 complexes, was almost 3-fold downregulated in T-PLL patients. Also, CDK4, CDK18, and CDK20 were upregulated 2-fold. It was therefore not surprising that the combined analysis of the microarray expression data with ex vivo drug response data from 12 patients revealed the cell cycle regulating the RB pathway to be upregulated and addicted in some of the patients. Furthermore, SNS-032 also inhibited the activity of NFAT, a central transcription factor and effector molecule of TCR signaling and T-cell activation. It has previously been demonstrated that T-PLL

mediating to a hyper-responsive and proliferative phenotype in the context of TCR engagement192.

Another group of compounds with a striking effect in T-PLL samples ex vivo were the p53 re-activator Prima-1 Met and MDM2 inhibitors serdemetan and nutlin-3. While

p53 mutations are frequently seen in many other cancers, in T-PLL these are not

commonly found193_{. However, the overexpression and accumulation of wild type p53}

is common in T-PLL, which could also be seen in our expression data. The efficacy of the p53 activators ex vivo in our T-PLL cohort indicates a new actionable pathway where these compounds act as an on-switch for the accumulated p53 and induce T- PLL tumor cell specific apoptosis.

In T-PLL, pathways such as JAK/STAT, PI3K/Akt/mTOR, and cell cycle signaling can all be dysregulated by different mutations and translocations. Therefore, it is not enough to simply abolish the effect of one of these pathways to hinder the progression of the disease. Combining different compounds allows for synergy or sensitivity that could not be achieved by the single compound alone. In the clinical setting, unless the compound acts on multiple pathways itself, it is essential to find the perfect

combination rather than relying on a single compound.

Overall, the CDK-inhibitor SNS-032 and p53 activators stood out as drug classes with high efficacy across nearly all T-PLL samples and should be further tested in this disease. The acknowledgement of the fact that successful ex vivo results are not a guarantee for a therapy with the desired effect in the patient is important. However, compounds that do not function well on ex vivo cells are unlikely to perform well in

vivo. As each round of ineffective treatment narrows the window of opportunity to

help the patient, it can be just as vital to avoid unnecessary therapy regimens. The concept of individualized system medicine could improve patient outcomes and reduce health care costs by eliminating these rounds of expensive therapy that have little effect on cancer but can severely diminish quality of life.

Conclusions and future perspectives

This thesis provides additional molecular information about the pathogenesis of T- LGL leukemia. The discovery of additional STAT3 mutations outside the mutation hotspot in the SH2-domain of STAT3 further highlights the importance of screening the entire gene for STAT3 mutations in LGL leukemia patients, as it is a viable diagnostic marker. In the rarer CD4+ cases, the high frequency of STAT5B mutations

presents with a novel diagnostic criteria alongside other established clinical and

hematological parameters. Additionally, in STAT-mutation negative patients, the STAT-activation seems to be mediated through heterogenous genetic events only visible through exome or whole genome sequencing. As sequencing costs decline even further, applying exome sequencing to every patient becomes plausible, which would further increase our knowledge about the genetic landscapes of many diseases. In T-LGLL, further studies are warranted to establish additional genetic events or

pathways that lead to T-LGL leukemia pathogenesis. A future task will be to develop

new molecules or therapeutic strategies targeting the frequently activated JAK/STAT pathway.

In T-PLL, ex vivo drug testing of primary patient cells has the potential to provide

novel personalized drug candidates such as SNS-032 and Prima-1 Met. These

observations hint at a centrally disturbed cell cycle regulation as well as a central role of TCR signaling in T-PLL. Despite high incidences of activating JAK/STAT

mutations and indications on pathway activations, inhibitors of the JAK/STAT pathway gave variable responses. Overall, ex vivo drug response patterns in T-PLL did not closely correlate with known genetic aberrations, suggesting that screening for recurrent genetic biomarkers cannot easily be turned into effective therapeutic

strategies in this problematic disease. This further emphasizes the importance of